[tex-live] Fixed: bug in the latest TeX Live snapshot.

[Tigran Aivazian]

Dear Peter and everyone else,

Thank you very much for fixing this bug. I will try to apply those patches 
in my local TeX Live copy and rebuild it and test if the PDF files now 
come out OK. If I can't then I will wait for things to be propagated into 
TeX Live's sources and do rsync/recompile from the official sources.

Kind regards
Tigran

On Tue, 12 Oct 2004, Peter Breitenlohner wrote:

> On Tue, 12 Oct 2004, Peter Breitenlohner wrote:
>
> In my mail earlier today I sent you the wrong files. Sorry. I had prepared
> all this on my notebook, from files that were missing one earlier bug fix.
>
> Here again the attachement, this times with the correct files.
>
> regards		Peter Breitenlohner
>
>>> On Sun, 10 Oct 2004, Karl Berry wrote:
>>> 
>>>>    My guess is that something goes
>>>>    wrong in the module <Handle a glue node for mixed...>,
>>>> 
>>>> Especially since that appears to be mostly new code in etex 2.2
>>>> (although bearing some resemblance to code in the do_glue function in
>>>> etex 2.1), seems like a good guess.
>>> 
>>> It is supposed to do the same thing as do_glue, but...
>>> 
>>>> Here is a simple file.tex, distilling from all the others that have been
>>>> posted:
>>>> 
>>>>    \output={\shipout\box255}
>>>>    \loggingall
>>>>    \TeXXeTstate=1
>>>> 
>>>>    \parindent=0pt
>>>>    \beginR\centerline{Vyr hVyryM}\endR
>>>> 
>>>>    \end
>>>> 
>>>> Processed with simply "etex file.tex".
>> 
>> Hi Karl,
>> 
>> you were almost right. The bug was in the reverse function that is used to
>> reverse all or part of the contents of an hbox in order to achieve
>> right-to-left typesetting. The bug was caused by not initializing the local
>> variable g_order; consequently some glue stetch/shrink was not accounted
>> for, thus the bad positioning of reversed text.
>> 
>> It is really a pitty that 'gcc -O2 -Wall' is almost useless due to the
>> amount of the produced noise, otherwise that would certainly have shown the
>> uninitialized variable.
>> 
>> The bug was most visible when reversing a complete box, but there might 
>> also
>> have been some subtle differences when reversing just a segment.
>> 
>> Karl, Thomas, Olaf, and everybody else,
>> 
>> attached is a file "etex-bugfix-2004-10-12.tar.gz" containing:
>> 1.	patch-24-etex-glue-rounding-bug
>> 2.	ChangeLog, etex.ch
>> 3.	etex.fix
>> 
>> (1) this patch contains the diffs to be applied to texk/web2c/etexdir.
>> Strictly speaking other people are not authorized to modify etex.ch; hereby
>> they have my permission to do so by applying this patch, provided they use 
>> a
>> patch program that properly sets the time stamps of modified files (such as
>> GNU patch with the option "-Z").
>> 
>> (2) these are the two modified files resulting from (1). Please keep the
>> timestamps!
>> 
>> (3) alternatively one could replace the empty file etexdir/etex.fix by this
>> one.
>> 
>> Olaf, Thomas,
>> 
>> can you please update the web2c and teTeX tarballs and see that things get
>> propagated to TeXlive.
>> 
>> regards
>> Peter Breitenlohner <peb at mppmu.mpg.de>
>
> Peter Breitenlohner <peb at mppmu.mpg.de>